LONDON – There has been much talk that Apple is going to drop Samsung as the manufacturer of Apple's own-designed ARM-based processor for its iPad, or at least for the next iteration, the A6. And TSMC has been nominated as the alternative, although Intel, with its recent opportunistic moves to get into foundry work has also been mentioned.
The falling out between Samsung and Apple at the system-level was often given as the reason why Samsung was history with Apple. Samsung is a rival to Apple with Android tablet computers and smartphones and the companies began trading patent infringement litigation in multiple locations. Apple initially sued Samsung in April, claiming that Samsung copied the look and feel of Apple products with its own smartphone and media tablet offerings. Samsung countersued Apple a few days later.
But Samsung was a system-level competitor to Apple before it was taken on as a chip provider so it does not necessarily follow that Apple will throw Samsung out – although that is still possible. Indeed the very fact that Samsung has taken Apple to court might suggest that Samsung thought the A6 commission was already gone and it had nothing to lose.
It has now been reported that TSMC is running the A6 through one of its fabs on a trial basis and that orders from Apple will depend on the yields achieved. Clearly Apple needs to satisfy itself that the number of wafers being committed to a process node will produce more than enough chips for its sales plans. But that also opens up other possibilities and begs the question as to what happens if yields are low or fall.
One possibility is that Apple could seek to qualify both Samsung and TSMC and operate them as mutual second-sources. Apple could then move orders between the suppliers to maintain downward price pressure.
After all, Apple already has multiple source suppliers of NAND flash memory. Indeed, some might argue that it is Samsung's ability to supply NAND flash memory has contributed to its success providing the current A4 and A5 processors to Apple.
Qualifying multiple sources does multiply the initial set-up costs. It is easier with standard product such as a memory and would not be undertaken lightly in the logic domain. In the past Apple has tended to use multiple sources of logic and mixed-signal chips in a serial fashion; designing a source of chips in for one generation of equipment and then giving the slot to a competitor in the next, but leaving the door open for the original manufacturer to get back in, if the product and the price are right. On the other hand the processor is the key component. The increased cost of double qualifying a processor might be compensated for by reduced unit costs and security of supply.
In addition, Apple might wish to apply lessons learned during the aftermath of the Japanese earthquake and tsunami of March 2011 by second-sourcing the A6 processor.
Apple was the world's largest chip buyer in 2010 and is predicted to spend about $22.4 billion on chips in 2011. Above all other companies Apple would have insight into the impact the earthquake had on the supply chain and the cost and desirability of true second sources.
Of course, that does not mean that Samsung is safe or even has a chance of supplying the A6. The same multi-source argument could apply to Intel and TSMC or even to all three companies. And Intel does have that rather attractive FinFET technology coming along.
The new Google Nexus S 4G phone is almost a carbon copy of the Samsung Epic 4G. Whether the S stands for Sprint or Samsung, I'm sure that Apple did not like Samsung supplying the latest Google phone... Apple has said in the past that Google is their greatest rival (not Microsoft) and that hone may have something to do with it.
Apparently Apple did/is using GF!
Just recently SAMSUNG increased it's production in SAMSUNG AUSTIN LSI FAB and offered FAB employees bonus to create additional production capacity and output to make up for GF failing to qualify Apple's A5 chip.
GF didn't exist when Apple starting running in Samsung, and I don't think the GF has nearly the 28nm capacity online -yet- that Samsung does. But other than that, it does seem like GF would be a much easier transition than TSMC, given that they predominantly share kits and technology with Samsung.
There must be a reason why Apple stayed away from GF. GF was quoted in several places having execution issues to bring technologies up and running in their Dresden Fab. Converting old AMD into foundry is not an overnight task especially FABs are spread around the globe. Also new GFmanagement are from AMD and they are still learning about what foundry is. You can't run foundry with OLD IDM mentality.
If I were Apple I would like to have at least two suppliers if not more (I worked at much smaller companies and we always made sure we had two suppliers of our chips)...so having both TSMC and Samsung sound ideal...Kris
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.